Background: The relationship between changes in proteinuria and myocardial infarction (MI) remains unclear in people with diabetes or pre-diabetes. We aimed to evaluate the predictive value and independent role of changes in proteinuria over a 2-year period in the incidence of MI in people with diabetes or pre-diabetes. Methods: Based on the baseline and 2-year dipstick screening results from the Kailuan prospective cohort study, participants were divided into four categories: no proteinuria, remittent proteinuria, incident proteinuria, and persistent proteinuria. Four multivariable Cox proportional hazard models were built to adjust for the effects of different confounding covariates. Results: Among the 17,625 participants in this study, there were a total of 238 incidents of MI during a median follow-up of 6.69 years. After adjusting for demography factors and laboratory indices, the association between persistent proteinuria and MI incidence was maintained (hazard ratio [HR] 2.50, 95% confidence interval [CI] 1.48-4.22). Every decrease of proteinuria from 2006 to 2008 was observed to be responsible for a 21% decline of MI incidence (HR 0.79, 95% CI 0.68-0.90). The interaction between changes in proteinuria and diabetes was confirmed with no effect on MI (P = 0.3371). Conclusions: Persistent proteinuria is an independent risk factor for MI incidence in the pre-diabetic and diabetic population. These findings may help clinicians to interpret proteinuria changes in the outpatient setting and provide possible preventive approaches for people with pre-diabetes or diabetes.